Reimagining overdue recovery: leading NBFC sees 211x ROI with Voice AI collections
Dvara KGFS
Key Statistics
- Number of attempted customers: 10.1K
- Number of uniquely reached customers: 6.4K
- Engagement Metrics
- Average call duration: 77 seconds
- % Promise-To-Pay the overdue amount: 28%
- % Promise-To-Pay Date : 21%
- Payment Conversion / Recovery Rate: 33% of reached customers
- ROI: 211x (INR 211 recovered for every INR 1 spent)
Client’s Challenge
In 2025, Awaaz AI partnered with the Dvara KGFS team to improve the overdue collections of INR 3 Cr (USD 334K) from 10.1K customers in the state of Tamil Nadu. These customers spanned multiple Days Past Due (DPD) buckets: 1-30 days, 31-60 days, 61-90 days, and 91-180 days. The target customer base spanned Tier 1-4 cities, towns, and remote villages of Tamil Nadu, making it cost-inefficient for field staff to reach everyone. Collection visits also reduce field staff productivity that could otherwise be directed toward business growth.
Traditional telecalling suffers from high fixed costs, frequent attrition, and variable agent quality, leading to inconsistent customer experiences. Convincing overdue customers to pay through an empathetic digital channel required specialized expertise. Based on Awaaz AI’s strong track record for human-centered design for microfinance and trust built over years through the performance of its IVR platform, Dvara KGFS team partnered with Awaaz AI to pilot an AI-enabled solution for collections with 10.1K customers. The goal was to validate its efficacy before scaling it to the complete overdue base of 50,000 customers.
Solution
Awaaz AI implemented a customized genAI voice agent in Tamil, fine-tuned on the customer’s transaction history, a use-case-specific workflow, and language/dialect nuances, ensuring each conversation felt contextual and natural. The agent was designed to be firm yet empathetic to the customer’s situation and reasons for overdue payments.
To deeply localize the solution:
- The agent’s conversational pathways and workflows were co-designed with Dvara’s team, keeping the target audience in mind. For example, highlighting the benefits of timely repayment while also explaining the negative impact of delayed payments on their credit score, helped create awareness among rural customers who initially declined to pay.
- Synthetic data and meta-prompting techniques ensured colloquial and consistent agent behavior across different Tamil dialects.
- The agent underwent continuous iteration, combining hill-climbing on conversion metrics, LLM-as-judge, human-in-the-loop QA, and auto-training mechanisms to improve accuracy and naturalness after each conversation.
Awaaz AI’s Impact
The campaign reached 10.1K overdue customers in Tamil Nadu, convincing 28% of the customers to clear the overdue amount and 21% to commit to a specific payment date within the next one week of voice AI calling. The pilot was designed without repeat calls to isolate the effectiveness of the AI agent.
At the end of the pilot, 2,131 customers (33%) had made the payment out of the 6,415 customers who were reached through voice AI calling, resulting in an overall collection of INR 63L (USD 70.5K), against the agent expense of INR 30K (USD 333), demonstrating an ROI of 211x. It is notable that the payment from non-reached customers from this pilot is almost zero, underscoring the effectiveness of Awaaz AI’s voice agents in generating additive collections. This impact has led Dvara KGFS to scale voice AI across their entire overdue base of 50,000 customers, reshaping their collections strategy and setting a new benchmark in the industry.
While most lenders traditionally depend on extensive field teams, Dvara KGFS embraced a blended AI-human model that elevates efficiency while preserving empathy. Dvara’s leadership showcased exemplary vision by adopting voice AI as a catalyst to fundamentally reimagine how overdue collections is done. At Awaaz AI, our mission is to help businesses and their customers connect through fresh, scalable, and intuitive voice experiences that define the future of customer engagement.
Testimonials
The level of localization and intelligence built into the solution was remarkable. The conversations felt natural, contextual, and culturally aligned - right down to Tamil dialect nuances and customer-specific repayment histories. The Awaaz AI team worked closely with us to design conversational workflows that genuinely resonated with our audience, helping customers understand the long-term value of timely repayments rather than simply pushing for collections. Awaaz AI has set a new benchmark for what technology can achieve in financial services, and we are excited to scale this partnership to redefine the future of customer engagement at Dvara KGFS.
Vincent Davit
Head of Direct (Customer Experience & Digital)